测试云联邦技术研发的时序异常检测数据集

时序异常检测数据集主要面向时序数据异常检测研究和相关算法优化需求建设，基于某平台内部服务器监控系统的日志和性能开源数据结合人工标注生成。主要记录了流量请求量、CPU使用率、内存占用率等时序观测值，数据量约为16.6MB。数据集包含真实业务场景中的时间序列数据，并人工标注了正常和异常点，覆盖多种异常类型（如突变、趋势变化、噪声干扰等），旨在为异常检测模型的开发、测试与评估提供高质量的基准数据支持。This time-series anomaly detection dataset is constructed to meet the research demands of time-series data anomaly detection and the requirements of relevant algorithm optimization. It is generated based on open-source logs and performance data from the internal server monitoring system of a platform, combined with manual annotations. It mainly records time-series observed values such as traffic request volume, CPU utilization, and memory occupancy, with a total data size of approximately 16.6 MB. The dataset contains time-series data from real business scenarios, with manual annotations for normal and abnormal points, covering multiple anomaly types such as mutations, trend changes, and noise interference. It aims to provide high-quality benchmark data support for the development, testing, and evaluation of anomaly detection models.

搜集汇总

背景与挑战

背景概述

该数据集专为时序异常检测研究和算法优化而构建，基于服务器监控日志与性能开源数据结合人工标注生成，包含流量请求量、CPU使用率等时序观测值。它覆盖多种异常类型，并提供真实业务场景中的标注数据，旨在支持异常检测模型的开发、测试与评估。